NSCT轮廓与主方向一致性红外与可见光图像配准

Infrared and Visible Image Registration Based on NSCT Contour and Main Direction Consistency

针对红外与可见光图像配准中特征点匹配难的问题,提出了NSCT轮廓提取与主方向一致性匹配的红外与可见光图像配准方法。先将NSCT变换用于提取红外与可见光图像的轮廓曲线图,然后使用CSS角点检测算法提取轮廓曲线图上的角点作为特征点,再以轮廓曲线的中线作为特征点的主方向,计算红外与可见光图像上特征点主方向的SIFT描述符作为特征点的特征描述,最后使用特征点主方向一致性匹配的方法去除误匹配的点,进而计算仿射变换参数。实验结果表明,所提方法的平均均方根误差为2.36,平均正确匹配率达到了91.6%,保证了两种图像相同特征点提取和匹配的准确度,从而提高了图像配准的精度。

Aiming at the difficulty of feature point matching in infrared and visible image registration,an infrared and visible image registration method based on NSCT contour extraction and main direction consistency matching is proposed.Firstly,NSCT transform is used to extract the contour curves of infrared and visible images,and the corners on the contour curves are extracted as feature points by using CSS corner detection algorithm.Then,the middle line of the contour curve is used as the main direction of the feature points,and the SIFT descriptor of the main direction of the feature points on the infrared and visible images is calculated as the feature description of the feature points.Finally,the mismatched points are removed by using the method of main direction consistency matching of the feature points,and the affine transformation parameters are calculated.Experimental results show that,the average Root Mean Square Error(RMSE) of the proposed method is 2.36,and the average Correct Matching Ratio(CMR) reaches 91.6%,which ensures the accuracy of extracting and matching the same feature points on two kinds of images and improves the accuracy of image registration.